Study Finds AI Chatbots Missed Hungary's Tisza in 90% of Voting Advice Tests
Updated
Updated · The Guardian · Jul 21
Study Finds AI Chatbots Missed Hungary's Tisza in 90% of Voting Advice Tests
3 articles · Updated · The Guardian · Jul 21
Summary
A Liberties study found ChatGPT and Gemini gave inaccurate, unstable voting advice in Hungary’s 2026 parliamentary election tests, often recommending the wrong party or ones not even on the ballot.
In the sharpest result, ChatGPT failed to recommend the Tisza party in 90% of detailed Tisza-aligned profiles and assigned it a percentage-match score in just 2% of cases.
The researchers ran five voter profiles 10 times each on both chatbots using direct-vote and percentage-match prompts; 96% of responses included parties not running, and identical prompts often produced materially different answers.
Fidesz-aligned profiles were recognized far more consistently, while both models often prefaced answers with disclaimers that they could not give political advice before delivering confident, persuasive recommendations.
Liberties said the errors likely reflect training-data gaps and language limits, but warned AI chatbots fall between the EU AI Act and Digital Services Act despite potential electoral risks in closer contests.
When an AI chatbot gives biased election advice, who is legally responsible for the resulting misinformation?
Are new global AI regulations strong enough to protect democracy from the influence of unreliable code?
How are AI-generated deepfakes already shaping election outcomes and eroding voter trust in political leaders?
AI Chatbots Undermine Hungary’s 2026 Election: How Algorithmic Bias and Disinformation Threaten Democratic Integrity
Overview
As Hungary heads toward its crucial 2026 parliamentary election, a study by the Civil Liberties Union for Europe highlights major flaws in leading AI chatbots like ChatGPT and Gemini. These chatbots, when asked for voting advice, failed to provide unbiased information, often omitting the opposition Tisza party led by Péter Magyar, despite polls showing it as a strong contender against Viktor Orbán’s Fidesz party. This bias in AI guidance raises serious concerns about the reliability of digital tools in shaping voter decisions during a highly competitive election, underlining the urgent need for greater transparency and accountability in AI systems.